The Vixen

Commercial amplifiers may not have been designed to be the best, but they were certainly designed to be the cheapest.

-- Myself

 

Vixen Development Prototype

Vixen Prototype

 

Specifications

Output: 30W / Channel; 8 Ohms
Frequency Range: 10Hz -- 37KHz (-3.0dbv; 8 Ohms)
Fixed Bias
Active Screen Voltage Regulation
Low Noise, Solid State Power Supply

 

Vacuum Tube Compliment (Each Channel)

1) 6J5

1) 6SL7

1) 6SN7

2) 807

1) 0C2

1) 6CB6

1) 6AQ5

 

Test Results

The following tests were conducted by subtracting the output signal from the input with a linear subtractor. The levels being adjusted for the deepest possible null at each frequency. The residual from the subtractor will be the distortion products. A perfect amplifier would produce a perfectly flat line since the output would be an exact copy of the input. Of course, no such amp has ever existed.

434Hz (1.0mS / 5.0mV)

434Hz

The base line was a bit jittery since the o'scope had difficulty locking on sync at this low level. This frequency provided the deepest null, indicating the least amount of phase shift. The residual distortion is mainly sinusoidal, as was the input signal. This indicates minimal harmonic distortion.

1.0KHz (1.0mS / 5.0mV)

1.0KHz

This signal is up to the 30mVP-P level. Here, you are beginning to get interference coming from a very noisy location (50KW AM BCB station 30 miles away). Still, a largely sinusoidal residual, indicating minimal harmonic distortion.

10KHz (0.1mS / 5.0mV)

10KHz

Some harmonic distortion is beginning to show up at the 10KHz frequency since the OPTs are beginning to lose effectiveness. This is to be expected, since no OPT is perfect across the entire audio band. Still no sign of extreme harmonic distortion that would greatly impair the sonic performance. What we do not find are any indications of cross over distortion, or artifacts that could be the result of instabilities.

For all these tests, the output was set to be just below the onset of clipping.

Design Philosophy

The following describes some of the corners we refused to cut.

The phase splitter used to derive the necessary 180° phase differential for the push-pull output stage is the differential amplifier. The differential (a.k.a. Long Tailed Pair) phase splitter has long been used in precision equipment such as laboratory quality oscilloscopes, and instrumentation amplifiers where accuracy is paramount. The LTP is preferred over phase splitters more commonly found in audio designs since it has, not only excellent phase-to-phase amplitude balance, but also balanced harmonic distortion over a wide range of frequencies. Amplitude and harmonic balance is further enhanced by the inclusion of an active, solid state CCS tail load. A solid state CCS has the advantages of higher impedance, and a more constant current than the usual solutions: a large passive tail load or a small signal pentode.

The grids of the finals have their own drivers. The inclusion of direct coupled drivers is rarely seen in commercial amps. A DC coupled driver overcomes the problem of capacitive coupling to the finals. During the overdrive that puts a positive voltage on the grids, the grid current, as the result of turning on the parasitic diode formed by the control grid and cathode, will charge the coupling capacitors to some higher negative voltage. (This is how Class C RF amps develop most, or all, of their grid bias). When the transient is over, the finals are biased closer to cutoff and operate with more distortion until the excess charge leaks off. With no coupling capacitor to accumulate a negative charge, a transient overdrive won't cause this distortion. Since the driver can source the grid current demand during overdrive, the finals can go into Class AB2 operation transparently.

Every text says that a Class AB1 vacuum tube with grids never driven positive requires drive voltage only, not current. This is not simply a gross oversimplification; it is just plain wrong. The grid has its input and stray capacitances; until that capacitance is fully charged, the grid does not see the true signal level. Charging any capacitance requires current. From the appearance of all too many designs, it is obvious that even experienced engineers have overlooked this detail. Small signal, high gain triodes simply do not have enough current sourcing capability to drive any load other than a high resistance, low capacitance one. This does not describe the grid circuit of very many power finals. If the amplifier can not follow a fast transient in the music, then it can not accurately amplify that music. This is the slew rate problem, and it does not just apply to solid state amps. A good, low impedance driver helps to greatly reduce this problem of inadequate slew rate, and the resulting distortion and loss of detail. Less clipping at the grids means better sonics at the ears.

It has been known since the mid-1930s that pentodes and beam "tetrodes" operate with greater linearity if the screen voltage is tightly regulated. To stiffen the screen voltage, an electronic voltage regulator is included. This regulator is far more accurate than the more common solutions: a bypassed voltage divider, or what's even worse: a simple series dropping resistor with a bypass capacitor. The electronic regulator also has a much lower AC impedance than any bypassed voltage divider or series dropping resistor. This eliminates a source of degenerative feedback that leads to increased harmonic distortion. The screen voltage remains fixed under conditions of increased screen current demand and power supply sag under heavy plate current conditions which occur when the inevitable high level transient hits the finals. Most commercial designs don't include voltage regulator tubes, much less an active regulator, to stabilize the screen supply. It is a major oversight.

Fixed bias gives much improved linearity and reduced THD over the more common cathode (auto) bias schemes. The spec sheets for all the audio power amp VTs show greater (sometimes much greater) THD figures for cathode bias than for fixed bias. Fixed bias also avoids a condition called "blocking" when rising plate current increases the voltage at the cathodes. In extreme cases, the amp can go silent as the finals are driven into cut off. Even if that doesn't occur, the finals will operate in a far less linear part of the plate characteristic until that excessive voltage leaks off. Cathode bias does not allow for transparent operation into AB2. This problem is made all that much worse by gNFB. Once the clip has occurred, the feedback loop loses control over open loop gain. This rising gain deepens the clip, forcing the finals further towards a cutoff situation. Add an intermediate stage with inadequate headroom (usually a cathodyne phase splitter doubling for a driver) that will add its own distortion to the mix, and the end result is truly a mess. Direct coupled drivers and fixed bias avoids all these problems, allows for more output power since none is wasted across a cathode bias resistor, and sounds better at high volume levels where transient overdrive is most likely to occur. Including a front end with a generous headroom margin (180VP-P for the Vixen) also reduces the distortion problem. With this design, the OPTs are driven into core saturation before the finals, or any intermediate stage, actually clips. However, there is a price to be paid: fixed bias isn't a set it and forget it proposition. The end user will have to take a bit of responsibility and actively manage his equipment. The reward is improved clarity and detail.

From the very beginning, the main idea was to produce a good sounding amplifier without negative feedback, either local or global, applied. In this manner, negative feedback is used the way it was meant to be used: to improve an already good sounding amp as opposed to being used to cover up a fundamentally poor design. This abuse of negative feedback occurs all too often. The end result has always been sound-alike amplifiers with a one-dimensional sound stage, flat bass, and no brightness to the mids and highs. Sometimes good enough "by the numbers" just isn't good enough.

Listening Impressions

Open Loop)

Running the amplifier open loop, it is obvious that we have succeeded in our main design goal: a very good sounding design as a foundation to build an even better amp. There are the expected problems that occur with pentode operation without corrective feedback. The bass performance is dominated by woofer resonance. Mids and especially the highs tend to be overly bright. With some program material, the high end had an unpleasant "stridency" or "shrillness" that quickly became hard to listen to for any extended time. This became most noticeable with Heavy Metal; with Classical, it was much less obvious. Even with no feedback at all, this amplifier already compares favorably with all but the best solid state designs, and beats a great many commercial amps with their lifeless sound.

Local Feedback Only)

After connecting the inner feedback loop, the improved damping has smoothed up bass passages noticeably. Improved bass is the first improvement noticed. The high end "stridency" that sometimes appeared with some material has been greatly reduced, but not eliminated completely. It is not so noticeable, and certainly does not produce the listener fatigue that the amp without any feedback produced. Playing one channel with the inner feedback loop and one without for comparison really shows the difference the inner feedback loop has made. We're already beyond most commercial amps so far as sonic quality.

The inner feedback loop has reduced considerably the third order harmonic during the Twin-T test to 6.0mVP-P, to 0.04% of VO. These tests verify the improved sonics due to adding this feedback loop.

Global Feedback)

Adding 20dbv of gNFB was absolutely horrible. The lifelessness of the sound was noticed at once. The detailing was exceptionally poor; the sound stage utterly flattened. Vocals seemed to fade into the bland background, and lyrics became harder to understand. This was worse in some ways than no feedback at all. The sound was quite solid state-like, and not good solid state either.

After reducing the gNFB to some 6.0dbv, the improvement over local feedback only is rather obvious: no more of that pentode harshness, and a sound that is not overly "aggressive". Bass packs plenty of "authority" -- felt as well as heard without excessive levels of volume -- without the remaining under-damped "sloppiness". Comparison listening with the most challenging program material demonstrates that the high end stridency has been eliminated. The detailing is excellent with subtle shifts of frequency and tone being easily heard. Vocals are crisp and clear. In playing programs with which one is familiar, it is amazing what can be heard that is missed when using many a commercial solid state product. With quality recordings -- analog as well as digital -- the Vixen does right by everything from Techno and Metal to Jazz and Classical.

About the 807

This particular vacuum tube began as the 6L6, developed in 1936 by O. Schade, of RCA. The 6L6 was designed to be an audio power amp. In this regard, it was quite successful. The 807 was developed as the RF version, having the then standard five pin base, a glass envelope, and a top cap to allow for operation at much higher voltages and power levels. A number of other types were also spun off the design. The 807W is a "ruggedized" version. The 1625 is the 12.6V heater version having an unusual seven pin socket, and the 6BG6 is the octal version of the 807, for use as a horizontal deflection amplifier for black and white television sets. All of these types will work in the Vixen design.


Copyright © 2009 All rights reserved